Investigating Internet Usage Behavior Modeling: A Multifractal Approach

Multifractal analysis has been implemented in many areas of research, such as network modeling, network intrusion detection, processing neural information, DNA sequence identification, turbulence modeling and many more. The success of this approach is because of its ability to identify persistent correlations on aperiodic, highly irregular time series. Our method is to collect real Internet usage data and then apply multifractal analysis techniques for each individual user. The results will be used as a feature vector for traditional machine learning techniques to identify a specific user given unlabeled Internet data. The proposed method could be applied to the aforementioned research areas plus applications of detecting legitimacy of network users, content delivery discrimination, theft alert, target tracking and other similar areas